Calculating Cancer Risk Through the Genetic Fingerprints of Tumors

Why do two people with the same cancer diagnosis—the same stage, the same cell type, and the same clinical profile—often have completely different outcomes? For decades, scientists have studied acquired mutations in cancer to find answers, but mutations alone explain only part of the story.

Now, a study published in Genome Medicine helps answer that question, uncovering that every cancer is shaped by a unique combination of inherited and acquired disruptions across biological pathways.

“Each cancer is as unique as the individual who suffers from it,” says senior author Lajos Pusztai, MD, DPhil, professor of medicine (medical oncology) at Yale School of Medicine. What drives the differences between patients, he says, is not the core set of pathways that all cancers share, but patient-specific disturbances that make each tumor biologically distinct.

In previous research, Pusztai's lab studied thousands of patients with cancer, and they saw a pattern—people who developed cancer younger tended to be born with more genetic vulnerabilities in cancer-relevant genes. However, those who developed cancer when they were older were born with fewer of those inherited vulnerabilities but had accumulated more DNA damage over their lifetime.

That inverse relationship raised a deeper question for his team: Could the combined effect of both inherited and acquired disruptions be measured together for each individual patient, and would it make each tumor biologically distinct?

To answer that, the research team built CanSys—an open access tool that produces a personalized biological damage report for each individual cancer.

At its core, Pusztai says, CanSys asks two fundamental questions about every gene in a patient's genome: How damaging is the mutation it carries, and how important is that gene to keeping a cancer cell thriving?

To figure out the first question, the team used a computational metric that assigns every possible DNA variant a damage score, predicting how severely it disrupts the protein that the gene produces. For the second they turned to DepMap, a platform that has collected data on the impact silencing individual genes has on the survival of a cancerous cell, across approximately 600 cancer cell lines.

Using these two measures, Pusztai and his team produced a gene-level impact score for every gene in a patient's tumor which were then assigned into cancer-relevant biological pathways and rolled up into pathway disturbance scores. The sum of these disturbances the researchers called the CanSys score.

While the gene-level score states how badly one specific piece of the cell's machinery is broken, the CanSys score quantifies the total damage to an entire biological system.

“Researchers can upload their data, and it can show you that in this particular individual cancer, what pathways are affected," says Pusztai, co-leader of the Genomics, Genetics, and Epigenetics Program at Yale Cancer Center. The tool is freely available at cansysplot.com.

When the team applied CanSys to more than 9,000 tumor samples across 31 cancer types in the Cancer Genome Atlas—the largest cancer genomics database—they found that a substantial proportion of cancer patients carried inherited disruptions in DNA repair, cell cycle regulation, and telomere maintenance pathways. These are the biological processes that keep genetic errors in check and regulate how long cells survive.

But when they ran the same analysis on 2,504 healthy individuals (people with no cancer diagnosis) from another open tool—1000 Genomes Project—the same inherited pathway vulnerabilities were present there too.

"Many of us are born with subtle pathway abnormalities," Pusztai says. "Each individual carries a different abnormality, but they converge at the pathway level."

The more inherited pathway disruptions a person has, the fewer additional changes their DNA needs before a cell becomes cancerous, the research team says.

"The more inherited impairment, the sooner people would actually develop cancer," Pusztai says. For now, this means that individuals with high inherited risk should carefully follow cancer screening recommendations and minimize exposure to carcinogens.

Pusztai's team is now seeking access to the UK Biobank—a dataset of approximately 500,000 individuals—with the goal of building a cancer risk score based on the totality of inherited pathway disruptions a person is born with. They aim to leverage the power of artificial intelligence to detect interactions between genes that older statistical methods simply could not capture.

"Why some individuals develop cancer is a mystery to so many people,” Pusztai says. “I really hope that in the next 10 to 20 years we are able to solve that mystery.”